Research map / Probabilistic and scientific modeling / Causal discovery
Addressing divergent representations from causal interventions on neural networks
Satchel Grant, Simon Jerome Han, Alexa R. Tartaglini, Christopher Potts
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add the Counterfactual Latent (CL) auxiliary loss to Boundless DAS | Boundless DAS on Llama-based LLM · Boundless DAS reproduction | EMD and IIA | Improved |
| train DAS with the CL loss instead of the behavioral loss only | DAS on synthetic MLP · Synthetic held-out classes | EMD: 0.032 -> 0.007 | Improved |
| train DAS with the CL loss instead of the behavioral loss only | DAS on synthetic MLP · Synthetic held-out classes (test) | IIA: 0.997 -> 0.9988 | No clear effect |
| train DAS with the CL loss instead of behavioral loss, evaluated on held-out partition | DAS alignment on synthetic dense/sparse partitions · Synthetic OOD transfer | OOD IIA | Improved |
| higher intervened divergence (training EMD) | DAS alignment · Synthetic OOD transfer | OOD IIA: coef. -.34, R^2=.73 | Hurt |
Closest papers by meaning
- The Non-Linear Representation Dilemma: Is Causal Abstraction Enough for Mechanistic Interpretability? (NeurIPS 2025 · similarity 0.91)
- Since Faithfulness Fails: The Performance Limits of Neural Causal Discovery (ICML 2025 · similarity 0.89)
- Everything, Everywhere, All at Once: Is Mechanistic Interpretability Identifiable? (ICLR 2025 · similarity 0.89)
- Causal Concept Graph Models: Beyond Causal Opacity in Deep Learning (ICLR 2025 · similarity 0.89)
- Neural Causal Abstractions (AAAI 2024 · similarity 0.89)
- Learning Linear Causal Representations from General Environments: Identifiability and Intrinsic Ambiguity (NeurIPS 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
